Leverage AI & Banking Data To Attract New Members & Optimize Banking Performance, TRANSCRIPTION

WEBINAR TRANSCRIPT

Leverage AI and Banking Data to Attract new Members & Optimize Banking Performance

Recorded September 17, 2024

Featuring:

Guest: Tom Tobin, CEO and Founder of Modelshop

Host: Brett Wooden, FI Strategist at Buildable

 

Brett: Welcome. This is kind of an experiment with Buildable, where we created this webinar series.  And really what we wanted to do is just connect you with some of the bright, intelligent minds out there in the fintech space and who are partnering with credit unions.  And so, I'm your host, Brett Wooden today and I'm really excited to have our guest, Tom Tobin, the founder and CEO of ModelShop. So welcome, Tom. 

Tom: Thank you, Brett. Thanks for having me. 

Brett: The idea with this webinar is Tom and I were having a conversation back at the World Credit Union Conference. And we talked about credit unions have this plethora of data, lots of data, and they're still trying to figure out, you know, what do we do with this data?  They've got a lot of vendors coming at them with ways that they can utilize the data.  And then you've got the other elephant in the room, right?  You've got this artificial intelligence that everybody is like, you know, do you need a strategy?  Do you have one?  How should you be implementing this in your credit union?  And so Tom and I thought, as we were having this conversation, you know what, let's jump on a webinar and just kind of have a fireside chat.  And Tom has agreed to have me pepper him with some questions.

And so really, that's where we're at today.  We're going to talk about leveraging AI banking data to attract new members and optimize banking performance. And so that's really the goal of our webinar. And so, Tom, I'm going to turn it over to you and start with the questions.  Tell us a little bit about yourself, but tell us more about what inspired you to create ModelShop.  And what are some of the key challenges you're aiming to solve in the financial space? 

Tom: Yeah, thanks. Thanks, Brett. So to start, I've been an automation geek since a very early age. It's funny, everyone talks about automation, but I was fascinated in seventh grade with this game that I saw called Labyrinths, where you kind of type and talk to it, and it responds back.  And originally, it was on the Apple II. So I started writing code, and I built a bunch of games that I loved sharing with my friends.  And I spent my whole career since then, you know, educational career and professional career, chasing that feeling, right?  The feeling of automation.  And everyone's talking about AI, but to me, it's the ability to do something that is so exciting to me.

So, I practiced this love in a few different areas, some really cool projects.  I worked for Callaway Golf really early on, and my degree was in machine vision. So I did a project where we would predict where the golf ball is going based on looking at the dimples on the golf ball. So, you know, fun, fun stuff. And then I never imagined I'd get into financial services. I didn't feel like it would be that exciting to me.

Then I found this small company called HNC Software in the mid-90s. Ultimately, they were acquired by FICO, but they were using neural networks to basically fix the fraud problem in credit cards. So very early use of that technology, what we refer to as AI now, and became enthralled with it. We built a bunch of great products there. We did the first real-time credit decisioning solution working with Capital One and Citibank. And, you know, I think we went live with that in the mid-90s. So, we'd go online and actually get a card number back instantly, which I thought was awesome. 

But in terms of how that turned into my entrepreneurial journey, I became frustrated with how companies were approaching analytic technology. My last role before starting ModelShop, I was at Fiserv, and I was leading their financial crime software products. And the key element that was missing, I think, is that most solutions separate the analytics from the decision. And they're treated as two independent things, different teams, different tools. And, you know, I saw that they had to come together.

So that's the genesis for ModelShop. It's model, but the shop is really the more important part of it. It's taking analytics and turning those analytics and data into actions that do things live, real-time, for your clients.

 

“The key element that was missing I think is that most solutions
separate the analytics from the decision.”

-Tom Tobin

 

Brett: Nice. Now, there's so many things to unpack there. So the first one is the game. So I was big into The Bard's Tale on the Apple II, which was kind of a similar type. And then you had mentioned, never thought you'd be in financial services. I think that's every credit union person, like when we were going to school, we were kind of doing our thing. Like, we never thought like, oh, let's go into the credit union industry. But that's really cool. So what makes ModelShop different? You know, there's a lot of AI driven companies out there, but what makes ModelShop different from other companies in the marketplace today? 

Tom: Yeah, a few things, you know, I did get frustrated. Every conversation about AI analytics, machine learning tends to come down to the math. And, you know, the math is important. And there's a lot of great tools that allow you to use historic data to build predictors, to do classifications. But the harder part is operationalizing that. 

Even if you think about self-driving cars, right? Everyone's, hey, it's AI, it's driving, you know, and a lot of people think about it as magic. And there is some math involved, absolutely, in creating a self-driving car. You know, you need to be able to use machine vision and predict, hey, there's a person walking in front of me. You need to model based on the behavior of the vehicle that's going to react to input. But in the end, 90% of that solution is automation.

It's bringing it all together to apply the gas, and to apply the brake, and to apply the steering. And it's engineering mixed with intelligence. And I think that was sort of what was missing.

So we start with data modeling. ModelShop is a very powerful data modeling tool. And it's because everybody's data is different. So we bring this data in, we blend it together. We draw inference from it. We allow you to build custom calculations, strategies, decision flows.

So I would say what makes us different – and then you layer on top of that, the analytics, whether it's other vendors that are providing analytics or models you build or predictive analytics you can build right in the platform – we bring it all together. It's really, we call it, the branding isn't ideal, but it's decision, orchestration, and optimization. Yeah, it's we're bringing it all together that last mile and helping you solve whatever problem you're trying to solve. 

Brett: That's awesome. I think one of our conversations too, and we probably should have asked you this right off the bat, but I love your definition of – when I asked you in a past panel – of what your definition of AI is, because I think a lot of people think AI is kind of a new thing and is just started. Do you mind telling us kind of a little bit, your definition? 

Tom: Yeah, I mean, it's been around for a long time. In the financial services world, we think about the FICO score. I worked for FICO for ten years after they acquired that company and worked on the technology behind the FICO score.

So we've been doing advanced machine learning, predictive analytics for a long time. And it goes way beyond, before that as well. So, you know, I think AI is anything that emulates what a human does effectively. And I think there is an aspect where AI has to take action. It has to do something. I mean, you can classify drawing six-fingered pictures of people as doing something. It is drawing. It's creating art to some degree. But, you know, it comes back to automation for me. It's achieving some goal. So, I have a pretty broad view on it. 

The new generative AI technology is really interesting because it personalizes the whole experience. It's like that human language interface.  And in a way, language is what makes us human, right? And that's what distinguishes us, our ability to communicate at that level. So allowing machines to do that is very exciting. But it doesn't replace everything else. 

I was at a conference just this week. And we had some really interesting conversations about the perception of AI taking over and saying, hey, I'm going to emulate what my loan officer is doing and sort of think through it and have an interaction with your member. And, you know, offer them options. But I really don't think that's how it's going to work.

I think it's going to be more the other aspects of AI that allow you to very quickly and interactively optimize products for your client and for your members and figure out their best strategy. The new generative AI will allow you to communicate that effectively and listen to them. And it makes it feel much more human, personal. It helps them feel like you know them as a member. 

But in the end, it's the same things in the background that are running through all the products and doing the ratios and figuring out, let's do an optimization, let's do a time-based projection of, you know, return on the assets that I'm going to lend to this member. That's the math that is all happening in the background. 

 

“The new generative AI technology…helps them [customers]
feel like you know them as a member."

-Tom Tobin

 

Brett: Yeah, oh man, the thing is, my background, I was a chief marketing officer and I remember the amount of data that we would have to manually go through to kind of find patterns and behaviors and then the products and services to market to members. And now how that's all just, I mean, kind of simplified.

The process was much more difficult back then. And I think too, the other one is, like I watched the Apple keynote and you're starting to see this kind of new generative AI be implemented in the mobile iOS devices. And I think that's going to be really interesting as well.

Tom: Just to add to that, you know, that's a great example because a lot of technologists I work with say gen AI is going to start writing code and that's not what's going to happen. AI is going to make it so you don't have to write code. 

Brett: Yeah.

Tom: The solutions are actually going to be able to interpret and think and come up with answers without a coding step. So we're not replacing coders, we're replacing code. 

 

“We’re not replacing coders, we’re replacing code.”

-Tom Tobin

 

Brett: Yeah, I agree a hundred percent. Yeah. I've got design friends who are kind of fighting back on the whole, you know, generative AI creating campaigns and working with Canva and all that stuff. And I kind of feel the same way, where it's not going to replace the designs themselves. It's going to help create those templates and things to help – especially the smaller credit union industry – marketing staff get the thing, and create it, and add their flavor to it. 

But, well, so when you look at ModelShop, how has it evolved since inception and what are some of the major milestones that your company's achieved? 

Tom: Yeah, it's pretty interesting. We've made a few adjustments along the way. When I started the company my vision was ultimately replacing the most powerful known code tool that exists. I don't know if you can guess what that is, but it's Excel. So, you know, we actually started the company with a focus on Excel-like models where, if you go to Wall Street and you go up to the 47th floor or whatever, and you look around, there's entire floors filled with analysts on spreadsheets, basically making banking work.

So that's how we started out. And we started with focusing on tier one organizations. We were actually part of the 2017 FinTech Innovation Lab in New York City, which was a great experience.

Met Jamie Dimon, met a lot of industry luminaries, mostly with a bit of a Wall Street focus. And it's sort of interesting. There were eight companies and two other companies you may have heard of that were my cohort members out of those eight – Alloy and Nova Credit.

Brett: Oh, nice. 

Tom: In a pretty heavy space there. But we ended up shifting away from tier one banks and that whole spreadsheet thing because there was sort of this, you know, cold, pry my spreadsheet from my cold dead hands, kind of thing that we're fighting against.

But where we really started to get traction and add value is when we went in and helped organizations solve complex problems that maybe they were using things like spreadsheets as a crutch, but what they really wanted to do was fully automate. And so, we started to really find our sweet spot in automation of sophisticated problems. We weren't limited to financial services.

We have a great healthcare client that's built care treatment pathway optimization models. So, we've solved some really interesting problems over the years, but as we're growing, what we're finding is more of a white glove – and that doesn't mean expensive. Or because we have this no-code platform we can move very quickly, but come in and say, you know, talk to the business and say, what problems are you trying to solve? Let's solve them in your terms, using your data, using your logic, and automate it, and make it part of your larger ecosystem.

 

“Let’s solve them [problems] in your terms, using your data,
using your logic, and automate it.”

-Tom Tobin

 

Brett: Yeah. So that's a great segue, so what are the core products and services offered by your company, ModelShop?  And then how do they help individuals in credit unions? 

Tom: Yeah. So, so we really have one platform, this decision optimization platform that can be applied to any problem set in financial services. A big part of our focus has been origination because I think that is one of the most critical early decisions that you make with members – who do we want to be our members and what kind of products can we offer them?  So, I think origination includes marketing product fit, right? Performance, it's all related.  And then it rolls right into servicing line management.

In my mind they all have to work together.  So, one of the challenges, you know, some of our clients have seen is – since they're in different parts of the business – there are different tools and those tools don't talk to each other.  And I think the mistake, I wouldn't say it's a mistake, but the first approach would be, hey, let's build a data warehouse.

Let's push all the data into one place so that we have a common view across our members and throughout their life cycle. And that's great. That's a good starting point. But what you really need is centralized member decision.  Decisioning is data plus logic, plus intent, plus options, all turned into an actual decision.  And, and that's what has to be made consistent across every member touchpoint.

So, we can address problems across the entire member life cycle.  And a lot of the solutions we create are bespoke to some degree, but we can do that very quickly.  And we have industry plugins, so we don't have to reinvent the wheel.

If you want, you know, if you want to start using Plaid data or you want to connect to the core or obviously bureaus or origination platforms, we are building a bigger and bigger set of plugins.  So very easily you get started; in 30 minutes you have sample data coming in from Plaid, for example, then it's a matter of how do we extend that to solve the problems we're trying to solve? 

 

“Decisioning is data, plus logic, plus intent, plus options
– all turned into an actual decision.”

-Tom Tobin

 

Brett: I think, you look at too with a credit union, there's, like you said, there's a lot of touch points and things where, if someone has access to all this, where would they start?  So how would they kind of work with you to find a starting place for that themselves or their credit union? 

Tom: Yeah, I always start with what problem you're trying to solve. That may sound a little rhetorical, but it's so important. Instead of saying we have a product out of the box, which I guess that we have plugins out of the box, we have starter sets, we have templates. You know, we really made it easy to come in and say, this is what a typical line management product would look like, or a marketing problem, product, or an origin, you know, credit origination product using banking data.  So here's a sample, but honestly, we really then say, okay, let's wipe the slate clean.

What problems are you trying to solve? And we'll pull in your data, which minimizes integration friction, right?  We don't have to translate things into our language.  And, you know, the first thing you want to do is do a proof of value or proof of concept.  And I typically recommend what we call models. Our models, what we call models, are really decision engines, analytic decision engines.  But let's get this model up and running in parallel as a first step and see how it performs.  First, let's replicate your existing logic.

That should be fairly straightforward.  But then let's start doing some tests and seeing what you might want to change.  I think part of this comes back to separating analytics from action and really trying to bring them together.

So, a key capability is almost anything you do, you can also run in a portfolio. So any variables, decisions, decision flows, it's very easy to say, okay, let's back up and run a hundred thousand members from the last year, or whatever, through this. And let's see how it would perform differently. And you can get swap sets, you can see differentiation between your slightly changing strategies.

 

“The first thing you want to do, is do a proof of value
 or a proof of concept."

-Tom Tobin

 

Brett: That's really cool. I like that. Especially even on the marketing side, I look at some of this stuff of trying to get deposit growth, loan growth, those type of things. And how do you target the specific members that you have and deepen those relationships? I could see utilizing this as well.

Tom: Yeah, and if your goal is to solve a specific product targeting a problem, for example, you're not sure you're getting the conversion that you expect from members signing up for a new product that you introduced or across your products. I mean, start there, start collecting that, take that data, feed it into a model that you create that starts on the analysis side. And then you can start to pull inference out of that. You can detect features.

We have true machine learning tightly integrated, where we're integrated with H2O and Python, scikit-learn algorithms. So you can do all that without coding and you can say, well, let's run a gradient boost analysis on this and see if we can pull out some common attributes and a predictor.

And then if you like that and you feel like there's value, then just deploy that one decision point that enhances your existing process. You don't have to rip everything out and do a big project to start.

Brett: Yeah, that's awesome. And another question, looking at your company, what do you think? This is one of my favorite questions to ask some of the fintechs too, that are in the credit union space or trying to get into the credit union space. But what do you see as your biggest challenges you've had at ModelShop? And how did you overcome them?

Tom: Yeah, I think the biggest challenge, well we've had a couple of challenges. First one is what I mentioned earlier, the whole, no one's going to let you replace their spreadsheet, or at least the current generation of people won't let you. I think future generations will be very open to it. But, you know, the second problem we ran into is just this whole data science is for data scientists.

And especially when we were working with tier one organizations, there's just a cultural organizational structure that you just can't bring people together. There's a bit of an ivory tower around the data scientists. They had to write a lot of code because writing code is important, I guess, to be a data scientist. And we had a hard time getting around that culture.

So, you know, we've continued to work to make it as simple and integrated as possible. You know, we may not quite yet be to the point where any business analyst can pick up the tool and build a whole automation from scratch. But that's where we come in and we help set it up, but then you can certainly maintain it. You can make changes to it. You can add new variables.

So, getting past that emotional connection to hardcore data science, where everything AI has to be some crazy complex machine learning algorithm. And they're there, but it's really 10% of the problem. The other 90% of the problem is automating this, connecting it into your systems, getting your products to line up, getting your pricing strategy to line up, adjusting downstream systems so you can handle dynamic pricing.

And that's been a tremendous challenge. I don't know if you know a whole lot about how some of these downstream, core systems, and account setups and, - you know, it's like you want more flexibility in your pricing strategy, but all the systems downstream sometimes have trouble handling it.

 

“Getting past the emotional connection to hard core data science, where everything AI has to be some crazy, complex machine learning algorithm."

-Tom Tobin

 

Brett: Yeah. It's interesting because, even talking about the products themselves, I was talking with a credit union yesterday and they're having a challenge trying to simplify their product offerings. And I think they, honestly have probably about eight checking account offerings, but then on top of that, they have their legacy checking accounts that they've offered in the past that they're essentially not sunsetting. And so that's where I could see something like this, they would definitely have to tackle that, what is their offerings? How is that going to connect? How do you find who, usage, all that stuff?

Tom: Yeah. And sometimes you have to run your existing products in parallel for a while and, you know, maybe you're not ready to move everything over at once. But, you know, that's where it's important.

I think that your decision automation platform has that flexibility. So, it's not uncommon for us to just suck in all the existing products in their entirety and, say these 37 products have to be replicated just as they are so that we can get everything flowing through a decision platform and here's where we're going to do our innovation. But it's where you need that platform flexibility to do that.

Brett: Right. Yeah, that's a great point. So share with us, I know you've got a case study out there as well, but share us some of the success stories of ModelShop and any highlights you have of users of ModelShop. We'd love to hear some of those stories.

Tom: Yeah. I think of every, especially many of our early customers, I think of them all as success stories because we help them, we love getting in and helping them solve a problem that would have taken a lot longer, a whole lot of code.

We can get them up and running very quickly. I'm very fond of one of our healthcare clients that we pretty much run the entire backend for. They were an early-stage company. They're growing and they're continuing to get investment, but their whole backend operation for this modeling, this healthcare treatment optimization modeling, is running on ModelShop models. So, I mean, that to me is the future. You could build a whole company, they built frontends and they have obviously the process, and there are scientists that are doing the logic, but they skipped the whole infrastructure part.

You know, they skipped writing all this custom code to run those models. But in the financial services space, probably the one I'm most proud of is – this is very early on in, I think it was 2015 or 2016 – where we worked with a sub-prime lender. And by the way, just a little side, we started doing it by doing a lot of work in sub-primes because they, in this financial services space, had very unique needs and they were trying to solve challenging problems.

My hope and belief is we can take a lot of those same technologies and that experience and help credit unions for good, right? You know, more equitable decisions, expanding their member base, competing with potentially predatory sub-primes. I'm not saying all are, but, where you can offer better alternatives if they have bad credit, or they maybe have run into some trouble.

So a little bit of an aside, but the customer was very, very sharp, ahead of risk. And they would basically underwrite every application with a full financial analysis of the person. So they had these crazy complex spreadsheets where they would do real cashflow projections, like time-based projections of potential probability of default based on a whole bunch of factors. And they were getting great results, but every one was create a new spreadsheet, get the data in there, and work it. So they were having trouble being competitive in their response times. And we came in and in four months, we completely automated what they were doing in the spreadsheet into an instant decision that was 15 milliseconds. And they went on to continue to grow and become a pretty, very established lender. So that's probably one of the ones I'm the most proud of.

Brett: Nice. I love it.

 

“In four months we completely automated what they were doing in the spreadsheet into an instant decision.”

-Tom Tobin

 

Brett: Yeah. And just because I referenced the case study, we'll send that out as well with the recording, so you can access that case study as well. And, before we open it up for questions, last question from me, Tom, what do you see for the future of ModelShop? You know, where are you going with all of this?

Tom: Maybe I'll answer that in the context of credit unions and members and how I think we can help impact that industry. Because I do see my ultimate vision is we can help lots of industries, but, you know, my background has been in financial services and that's where a lot of our growth is right now. So we're very focused on that. And I think there's an opportunity to help drive increased equity, the ability to compete for credit unions.

You know, the big elephant in the room is how do you attract new next generation members? And I am a hundred percent convinced that millennial and Gen Z sons, that if you can't provide an instant decision, you're not going to have members, right? It's just becoming more and more clear. So, I think our biggest success metric would be helping get lots of credit unions to very high levels of automation with a frictionless environment – and we're not solving the whole thing.

We work with partners and there's the frontend experience, and there's backend set up and the origination platform. But we can be the brain behind that, where we can take exactly what you're trying to do with your member relationships, what culture you're trying to exhibit, trying to make those members feel like you know them.

And you do know that if you just happen to know them through technology, that's good. It's an extension of you as a credit union culture. So helping enable that. And that's why I think, you know, credit unions are so interesting to me. So high degree of automation and increased equity in decisions. And to me, I think of that as not just totally leveling the playing field for everybody. I mean, that's part of it, but what's really exciting to me is finding people who maybe had been adversely impacted by a couple of things.

One is using kind of stale perceptions of financial behavior and penalizing people that don't exhibit those exact behaviors. You know, maybe they don't leverage credit a lot or maybe they just think about their financial relationships differently, which is exhibiting a lot of the features of the younger generation, as well as helping those who run into financial trouble and need help.

So if a credit union employee, right, can sit across the table and hear the whole story and help that member, right? That's wonderful. It's what makes credit unions great. But if you can't scale that and automate it, then it's going to be a loss of power. So, capturing that art and turning it into intelligence so that you can do the same thing for your members, but in real time at two in the morning, when the alpha is finishing up their schoolwork and needs to get a loan, that sort of thing.

 

“The big elephant in the room is how do you attract
new next generation members.”

-Tom Tobin

 

Brett: Yeah, it's funny. I was a chief operations officer at a healthcare credit union, and we saw a huge influx of our loans when I think it was like the 1:30 AM shift got off. You would see this influx, but it would go into our queue. So, you know, there wouldn't be decisions until we got in the office the next morning at 7:30 AM. Yeah, so I definitely, I'm hearing you on a lot of this and even having a daughter going to college, you know, we got her a credit card through a credit union, but it was a process. It was filling out an application, getting put in a queue to review, and then we had to get on a phone call where she went to the Apple card and it was three clicks that she was able to get an Apple card through her phone.

So yeah, I will say it is exciting times because we are looking at the member experience and how can we simplify that and make it more efficient. And I think tools like this help as well.

And we did have exactly 13 attendees. I am a numbers person. So, on Friday the 13th, we have 13 attendees.

Bottom of Form

Well, it looks like we don't have any questions. First, we really want to thank you for joining this webinar today. Really the goal is just to educate the folks and highlight the bright minds of the fintech space, the credit union space. And so we'll be doing a follow-up email. In that follow-up email we'll have the recording, we'll have how you can access the white paper, and then how you can get ahold of Tom or myself, if you have any additional questions, or kind of want to see it in action and hear more, some of the success stories. And then Tom, are you going to be at any other conferences the remainder of the year?

Tom: I don't think so. I think we're kind of wrapping up for the year, but yeah, we've been part of some great ones this year. We're enjoying really getting more established in the credit union space, so looking forward to next season.

Brett: Yeah. Great. Well, again, everybody have a great Friday. Thank you all for joining. And yeah, look out for that email with all the additional information and thank you again, Tom, for joining.

Tom: Yeah. I appreciate it, Brett. Everybody have a great weekend.

Brett: All right. Thank you.

 

End Interview

 

Thank you for joining us!